Skip to main content
Open Access Publications from the University of California

Design of a maintenance and operations recommender


We describe the design of a maintenance and operations recommender. The recommender uses information from computerized maintenance management systems (CMMS) and energy management and control systems (EMCS) to recommend what maintenance personnel should do in response to a maintenance service request or other event requiring a maintenance or control system action. The recommender integrates text information from a CMMS database and sensor information from an EMCS to provide recommendations. Text is processed using the Extended Boolean model, which is a simple text processing method commonly used to retrieve information from large databases. The recommender compares text problem descriptors and sensor data descriptors to estimate the similarity between previous maintenance actions and the maintenance action that will be taken. Actions with a high predicted similarity are more highly recommended than those with a low predicted similarity. The recommender uses reported maintenance actions to learn to improve its recommendations. It compares predicted similarity indexes with similarity indexes computed by comparing maintenance actions. The difference between predicted and computed similarity is used as an error signal to adjust weights that relate similarity indexes of individual problem descriptors. We use a simple example to demonstrate the steps that the recommender uses to make recommendations and to learn.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View